time-nuts@lists.febo.com

Discussion of precise time and frequency measurement

View all threads

Re: [time-nuts] 1PPS users?

HM
Hal Murray
Mon, Dec 19, 2016 12:36 AM

So, if an NTP user can get his time fix directly from a noisy receiver, who
actually needs a time-accurate, low jitter 1PPS pulse?

Most kernels have an option to capture a time stamp from a PPS signal at
interrupt time.  That is much more accurate than the timing you get from user
mode on a serial data stream.

There is also a mode where the whole NTP PLL processing is done in the
kernel.  I don't see why that should make as much of a difference as it does,
but I haven't tracked down the details.  (It's not in the typical Linux
kernel.  You have to build your own.)

Most low cost GPS receivers have crappy timing on the serial port.  Really
crappy.  It wanders with a time scale of hours so you can't filter out the
jitter by averaging for a minute or two.  PPS on that sort of unit makes it
much more interesting.

--
These are my opinions.  I hate spam.

bob@evoria.net said: > So, if an NTP user can get his time fix directly from a noisy receiver, who > actually needs a time-accurate, low jitter 1PPS pulse? Most kernels have an option to capture a time stamp from a PPS signal at interrupt time. That is much more accurate than the timing you get from user mode on a serial data stream. There is also a mode where the whole NTP PLL processing is done in the kernel. I don't see why that should make as much of a difference as it does, but I haven't tracked down the details. (It's not in the typical Linux kernel. You have to build your own.) Most low cost GPS receivers have crappy timing on the serial port. Really crappy. It wanders with a time scale of hours so you can't filter out the jitter by averaging for a minute or two. PPS on that sort of unit makes it much more interesting. -- These are my opinions. I hate spam.